An adjusted pay gap estimates the remaining difference in pay between women and men after selected explanatory factors are taken into account. Employers may use job, grade, location, working time, tenure or other defensible variables to investigate why an observed gap exists. The method can be useful, particularly in larger datasets, but Directive (EU) 2023/970 does not prescribe one universal adjusted-gap model and Article 9 reporting metrics remain separate. An adjusted result should therefore be treated as evidence for investigation, not as an automatic legal conclusion that a difference is justified or unjustified.
Jurisdiction: European Union
Adjustment Tries to Compare More Similar Records
An unadjusted result reflects all differences embedded in the observed workforce. An adjusted analysis attempts to hold selected characteristics constant so the analyst can estimate whether a pay difference remains after those characteristics are considered. For example, an employer may examine job, grade, location and working time when those variables are relevant and reliably measured. The objective is not to manufacture a smaller number. It is to understand which measured factors account for part of the observed disparity and which differences remain unexplained by the selected model.
The Choice of Control Variables Is a Substantive Decision
Every control variable embeds an assumption about what should be treated as a legitimate explanatory factor. A model that controls for grade may be reasonable if the grade structure itself is objective and gender neutral. The same variable may be problematic if historic promotion or classification practices produced biased grade outcomes. Similar questions arise with performance ratings, prior salary, tenure and experience. Analysts should therefore document why each variable is included, how it was measured and whether using it could absorb the effect of a process that itself deserves equality review.
Adjusted Results Do Not Replace Article 9 Reporting
Directive (EU) 2023/970 specifies observed reporting metrics, including the gender pay gap, median gender pay gap, variable-pay gaps, quartile distributions and category-level information. The Directive does not say that an employer can replace those figures with a regression-adjusted result. Adjusted analysis can sit beside the required metrics as an investigative layer. This distinction is important for governance because published compliance figures, internal diagnostic models and legal equal-pay assessments serve different purposes and should not be combined into one number without explanation.
A Smaller Adjusted Gap Does Not End the Review
If an observed gap becomes smaller after controls are added, the model suggests that measured differences in workforce characteristics explain part of the original disparity. That finding is useful, but it does not automatically show that the underlying processes are fair. The employer may still need to ask why women and men are distributed differently across grades, roles, bonus opportunities or locations. It should also examine same-work and work-of-equal-value categories where individual or group differences remain. Adjustment can clarify the analytical picture, but it cannot replace organisational and legal judgement.
A Residual Adjusted Gap Requires Careful Interpretation
A remaining adjusted gap may indicate that the selected variables do not fully explain the pay difference. It may also reflect missing data, measurement error, omitted variables, model specification or small sample effects. Statistical significance can help assess uncertainty, but it is not the only relevant consideration for equal-pay compliance. A practically meaningful difference in a small worker category can deserve review even where a conventional significance threshold is not met. Employers should combine model results with pay records, job evaluation, compensation decisions and other evidence before deciding on remediation or legal conclusions.
Keep the Adjusted Method Reproducible
A defensible adjusted analysis should record the population, dependent variable, group indicator, control variables, transformations, exclusions, model specification and software version. Analysts should preserve both the unadjusted result and each important adjusted specification rather than retaining only the final preferred model. Sensitivity checks can show whether conclusions change materially when reasonable variables or data treatments change. This audit trail is especially important when a model informs pay corrections, management decisions or legal review because stakeholders need to understand how the result was produced and what assumptions it depends on.
Frequently Asked Questions
What is an adjusted pay gap?
It is an estimated pay difference after selected explanatory variables are taken into account, often using regression or another multivariable method.
Does the EU Pay Transparency Directive require employers to report an adjusted pay gap?
The Directive specifies its own reporting metrics and does not prescribe one universal adjusted-gap model as a replacement for them.
Can an adjusted gap prove that pay differences are lawful?
No. It can provide useful evidence, but legal justification depends on the relevant equal-pay framework, objective gender-neutral criteria and the facts of the comparison.
Related Guides
Official Sources
Requirements and practices differ by jurisdiction and organisation. Check current local law, official guidance and professional advice for a specific situation.